Research On Ridgelet-based Motion Compensated Video Codec

With the emergence of communication networks, along with the rapid growth of multimedia service, the role played by multimedia communication has become increasingly important and essential in next generation communication system. Therein, video coding is the fundamental and kernel technology in emerging multimedia video communication. In the communication environment, transmission bandwidth is limited. To transmit video over a limited transmission channel with video quality un-degraded, video codec need higher coding efficiency than before. We focus on the key technologies of inter prediction coding in this dissertation,the main works of mine are the following:First of all, a brief review of video coding theory, history and development of video coding standard is also offered, then deeply research into the basic theory and search algorithm of motion estimation. The paper presents a novel fast search algorithm for block base motion compensation called the Orthogonal Diamond Search, which consists of two sequential search stages: 1) adaptive orthogonal initial search and 2) refined local search. The performance of the algorithm is evaluated by using MPEG test sequences and the results are compared to the traditional well-known foll search algorithm, diamond search algorithm and the three-step search algorithm.Then,introducing the major theory of wavelet and deeply researching into the image compression algorithms such as EZW, SPIHT. The paper uses the Claire as the test image and compares the CR, PSNR, Encoding Time of DCT and SPIHT under different threshold.Finished, a brief review of ridgelet theory, including a description of a finite, discrete algorthm is offered. We focus on the Ridgelet-based Motion Compensated Video Codec, and propose a method composed by two stages: first the FRIT is used to represent lines with biggest coefficients and use Zero Run-Length Coding compress biggest coefficients, afterwards the reconstructed image is subtracted from the original one and the residual image can be coded with SPIHT.The results show that the strength of the Orthogonal Diamond Search lies in it drops down the number computations required per macro block by more than five times to the three step search algorithm. Furthermore, the accuracy of the orthogonal diamond search algorithm is comparable to that of the three-step search or diamond search, so it benefits to get residual image. Experimental results also clearly show that the proposed compression technique result in higher quality reconstructed images compared to Wavelet-based compression algorithm operating at similar very lower bit rates for the class of residual image where edges are dominant.